10 lesser-known facts about ChatGPT

1. It doesn’t “think” like a human. It predicts the most likely next words based on patterns in data rather than reasoning the way people do.
2. It can sound confident while being wrong. This is called a “hallucination.” Plausible-sounding answers aren’t always accurate, especially for niche or recent topics.
3. It has no memory by default across chats. Unless memory is enabled and you’ve chosen to save information, it doesn’t remember previous conversations once they end.
4. The wording of your prompt matters a lot. Asking for a role (“Act as a lawyer”), constraints (“Use bullet points”), or examples can dramatically improve results.
5. It can explain its reasoning, but not reveal its internal reasoning process. The explanations you see are summaries, not the model’s internal computation.
6. It performs surprisingly well at many tasks because it’s general-purpose. It can write, code, translate, summarize, brainstorm, tutor, and analyze—all using the same underlying language capabilities.
7. It benefits from iteration. Instead of expecting a perfect first answer, refining the prompt or asking follow-up questions usually leads to much better results.
8. It doesn’t automatically know current events. For up-to-date information, it needs access to current web sources or other live data tools.
9. It can adapt its style. You can ask it to write like a teacher, editor, scientist, comedian, or in a specific reading level or format.
10. Its biggest strength is collaboration. Rather than replacing human judgment, it’s most effective as a brainstorming partner, editor, coding assistant, research helper, or tutor.

A bonus “power user” tip: instead of asking “Explain quantum computing,” ask “Explain quantum computing as if I’m a 12-year-old, using analogies, then give me three real-world applications and a short quiz.” The more specific your prompt, the more useful the response tends to be.